An adaptive test against ordered alternatives

Nonparametric tests are presented for the c-sample location problem which are efficient for detecting an upward trend in the treatments. In this paper, only linear rank statistics are considered to perform the test. While the choice of suitable regression constants depends on the test problem, the choice of scores depends on the underlying distribution. Thus, a distribution-free adaptive test is proposed which takes information from the data into account. The adaptive test is compared with the test of Jonckheere (1954) and Terpstra (1952) via Monte Carlo simulation. (C) 1997 Elsevier Science B.V.